Hybrid DEM-LiDAR UGV planner hits 95% obstacle avoidance, cuts slopes to 2.7°

A team led by Devender Singh and colleagues has published a new framework for autonomous ground vehicle navigation that bridges the gap between coarse global maps and real-time local sensing. Their system uses a Digital Elevation Model (DEM) to precompute a global path with the A* algorithm, then relies on LiDAR data for adaptive local corrections. This hybrid approach lets UGVs react to sudden environmental changes—like rocks, ditches, or steep grades—without losing sight of the overall route.

In Gazebo simulations across custom terrains, the framework beat a baseline by achieving a 95% obstacle avoidance rate, while cutting the average slope the vehicle had to traverse from 8° down to just 2.7°. That reduction means the UGV picks safer, more energy-efficient paths in real time. The work demonstrates how fusing low-cost global data with onboard sensors can deliver robust performance in dynamic outdoor settings. It points toward more reliable autonomous navigation for applications like agriculture, mining, and military logistics, where terrain unpredictability remains a major hurdle.